451

A p p e n d i x F

TIME-RESOLVED MOLECULAR CHARACTERIZATION OFORGANIC AEROSOLS BY PILS + UPLC/ESI-Q-TOFMS

Zhang, X., N. Dalleska, D. Huang, K. Bates, A. Sorooshian, R. Flagan, and J.Seinfeld (2016). “Time-resolved molecular characterization of organic aerosolsby PILS + UPLC/ESI-Q-TOFMS”. In: Atmos. Environ. 130, pp. 180–189. ���:http://doi.org/10.1016/j.atmosenv.2015.08.049.

AbstractReal-time and quantitative measurement of particulate matter chemical compo-

sition represents one of the most challenging problems in the field of atmosphericchemistry. In the present study, we integrate the Particle-into-Liquid Sampler(PILS) with Ultra Performance Liquid Chromatography/Electrospray ionizationQuadrupole Time-of-Flight High-Resolution/Mass Spectrometry (UPLC/ESI-Q-TOFMS) for the time-resolved molecular speciation of chamber-derived secondaryorganic aerosol (SOA). The unique aspect of the combination of these two well-proven techniques is to provide quantifiable molecular-level information of particle-phase organic compounds on timescales of minutes. We demonstrate that the appli-cation of the PILS + UPLC/ESI-Q-TOFMS method is not limited to water-solubleinorganic ions and organic carbon, but is extended to slightly water-soluble speciesthrough collection e�ciency calibration together with sensitivity and linearity tests.By correlating the water solubility of individual species with their O:C ratio, a pa-rameter that is available for aerosol ensembles as well, we define an average aerosolO:C ratio threshold of 0.3, above which the PILS overall particulate mass collec-tion e�ciency approaches ⇠0.7. The PILS + UPLC/ESI-Q-TOFMS method can bepotentially applied to probe the formation and evolution mechanism of a variety ofbiogenic and anthropogenic SOA systems in laboratory chamber experiments. Weillustrate the application of this method to the reactive uptake of isoprene epoxydiols(IEPOX) on hydrated and acidic ammonium sulfate aerosols.

F.1 IntroductionChemical characterization of particulate organic compounds is crucial to un-

derstand the formation and evolution of secondary organic aerosol (SOA). The

452

conventional method for particle-phase molecular speciation is to collect aerosolson a filter substrate, followed by extraction and preconcentration. Analysis of filterextracts is commonly performed by liquid chromatography (LC) and gas chromatog-raphy (GC), coupled to mass spectrometry with the use of electron ionization (EI),chemical ionization (CI), electrospray ionization (ESI), and atmospheric pressurechemical ionization (APCI) (Dye and Yttri, 2005; Lavrich and Hays, 2007; Linet al., 2012; Simpson et al., 2005; Surratt et al., 2007a; Surratt et al., 2007b; Surrattet al., 2006; Szmigielski et al., 2007). Major organic classes in SOA that have beenidentified from filter-based analysis include (nitrooxy)-organosulfates (Chan et al.,2011; Iinuma et al., 2007; Surratt et al., 2007a; Surratt et al., 2008, 2007b), dimers,trimers, and oligomers (Gao et al., 2004; Jang et al., 2002; Kalberer et al., 2004;Limbeck et al., 2003; Schilling Fahnestock et al., 2015), and humic-like substances(Gelencser et al., 2002; Graham et al., 2002). A limitation of filter-based analysis islow time resolution and, consequently, the inability to track particle-phase kinetics.In addition, filter artifacts, such as adsorption of ambient vapors and evaporation ofsemi-volatile compounds from the filter surface, lead to uncertainties in the quan-tification of particle-phase components (Dzepina et al., 2007; Schauer et al., 2003;Turpin et al., 2000).

Powerful, on-line techniques have been developed for chemical speciation of or-ganic aerosols, with rapid time response and minimum sample handling. The generalprinciple is to vaporize airborne aerosols by thermal or laser desorption, followedby ionization and mass spectrometric detection (Canagaratna et al., 2007; Sullivanand Prather, 2005). The Aerodyne Aerosol Mass Spectrometer (AMS), which isnow a routine component of ambient and chamber studies, enables measurement oforganic fragments and derivation of the atomic O:C and H:C ratios (Aiken et al.,2007; Aiken et al., 2008; Jimenez et al., 2003). Identification of individual speciesby AMS is not available due to the high evaporation temperature (600 �C) and EIenergy (70 eV), which result in significant molecular fragmentation. Moreover, allthermal/laser desorption methods are susceptible to fragmentation of non-refractorycompounds. To achieve unambiguous molecular identification of particulate organiccompounds, extensive fragmentation is avoided by employing soft ionization tech-niques such as chemical ionization (Aljawhary et al., 2013; Hearn and Smith, 2004;Hellen et al., 2008; Ho�mann et al., 1998; Lopez-Hilfiker et al., 2014; Smith et al.,2005; Yatavelli and Thornton, 2010), photoelectron resonance capture ionization(LaFranchi and Petrucci, 2006; Zahardis et al., 2006), and vacuum ultraviolet singlephoton ionization (Ferge et al., 2005; Isaacman et al., 2012; Northway et al., 2007;

453

Oktem et al., 2004). Full-scale implementation of these methods into routine mea-surements requires achieving short measurement cycles, resolving quantificationcapability, interpreting the mass spectral complexity, and comparing data recoverywith conventional methods.

The Particle-into-Liquid Sampler (PILS), first developed by Weber et al. (2001),grows aerosols by supersaturated water vapor condensation, producing dropletssu�ciently large for collection by inertial impaction. The resulting liquid samplescan be analyzed by ion chromatography (IC) for water-soluble inorganic ions andsmall dicarboxylic acids (Ma et al., 2004; Oakes et al., 2010; Orsini et al., 2003;Rastogi et al., 2009; Sorooshian et al., 2008; Sorooshian et al., 2007), or by a totalorganic carbon analyzer (TOC) for the total water soluble organic carbon (WSOC)(Peltier et al., 2007; Sullivan et al., 2006). PILS coupled with mass spectrometry hasalso been deployed for the analysis of WSOC in field and chamber measurements.Bateman et al. (2010) compared the o�-line mass spectra of limonene + O3 derivedSOA samples collected by PILS with those collected on filter substrates and foundthat the peak abundance, organic mass to organic carbon ratios, and the average O:Cratio are essentially identical. Parshintsev et al. (2010) integrated PILS on-line witha solid-phase extraction chromatographic system for the characterization of organicacids. Using a similar concept, Clark et al. (2013) directly injected PILS samplesinto the ionization source of the mass spectrometer and validated the PILS-ToF-MSsystem against other particle measurement methods in terms of total ion abundanceand average O:C ratios of isoprene and ↵-pinene derived SOA.

A primary challenge in the deployment of PILS as an e�ective particle collectiondevice lies in the determination of the overall mass collection e�ciency. Ambientand chamber derived organic aerosols usually comprise of thousands of specieswith various physicochemical properties. Since PILS was originally designed touse water steam as the growing agent, one expects that particles with an extremehydrophobic surface would be di�cult to collect. Therefore, a guideline needs tobe developed for the reference of PILS selectivity to certain types of aerosols. Thisis one main focus of the present study.

Here we combine PILS with Ultra Performance Liquid Chromatography/ Electro-spray Ionization Quadrupole Time-of-Flight High-Resolution Mass Spectrometry(UPLC/ESI-Q-TOFMS), for time-revolved molecular-level characterization of SOAduring chamber experiments. This technique is particularly suited to polar or water-soluble organic molecules, and potentially high molecular weight species, such

454

as organosulfates. Advantages of this technique include: 1) No need for samplepreparation; 2) Molecular-level information can be inferred from the electrosprayionization process, in which ions are, in general, observed as those of the parentmolecule with the addition of an H atom (positive mode) or removal of an H atom(negative mode); 3) Temporal profiles of particle composition can be obtained ona time scale consistent with that of SOA evolution; and 4) Chromatographic sep-aration of organic compounds with di�erent polarities avoids the analyte signalsuppression that occurs during the electrospray process of organic mixtures, leadingto molecular-level quantification of particle-phase constituents. We demonstratethat the PILS + UPLC/ESI-Q-TOFMS method is suitable to measure water-solubleorganic carbon (WSOC), as well as less hydrophilic or slightly water-soluble com-pounds. A collection e�ciency of >0.6 can be achieved for chamber-derived SOAsystems with average O:C ratios >0.3. We illustrate the application of this tech-nique to the reactive uptake of isoprene epoxydiols (IEPOX) on hydrated and acidicammonium sulfate aerosols.

F.2 Experimental MethodsF.2.1 Pure Component Organic Aerosols

Table F.1 gives the chemical properties of organic standards (purchased fromSigma Aldrich) that were used to generate pure component organic aerosols. Thesestandards include carboxylic acids, amines, and polyols that span a broad range inwater solubility from miscible to insoluble, and vapor pressure from intermediate tolow volatility. Each standard was dissolved in water or isopropanol to produce a con-centrated solution. Aerosols with known chemical composition were produced byatomizing each single concentrated solution followed by desolvation. Two di�usiondenuders filled with silica gel and activated carbon were used to remove the solventprior to injection into the Caltech 24 m3 Teflon chamber. Relative humidity (RH) andtemperature in the chamber were maintained at <5% and 25 �C, respectively. Thesize distribution and number concentration of the pure component organic aerosolswere measured continuously using a custom-built scanning mobility particle sizer(SMPS) consisting of a di�erential mobility analyzer (DMA, TSI, 3081) coupledwith a condensation particle counter (CPC, TSI, 3010). More details of the SMPSoperation can be found in Loza et al. (2014), Zhang et al. (2014a), and Zhang et al.(2014b).

455

chemical exact m/z water solubility vapor pressurecompound formula mass (error/mDa) at 20 �C (atm at 298 K)Adenine C5H5N5 135.0545 [M+H]+ (1.6) 0.103 g L�1 8.117 ⇥ 10�5

Adonitol C5H12O5 152.0685 [M-H]� (4.8) 50.0 g L�1 7.711 ⇥ 10�12

Adipic acid C6H10O4 146.0579 [M-H]� (0.6) 0.25 g L�1 3.292 ⇥ 10�9

d-Sorbitol C6H14O6 180.0634 [M-H]� (2.5) 182.0 g L�1 2.424 ⇥ 10�14

Diethylmalonic acid C7H12O4 160.0736 [M-CO2-H]� (0.1) 25.0 g L�1 2.696 ⇥ 10�9

Vanillic acid C8H8O4 168.0423 [M-H]� (-1.8) 1.5 g L�1 7.940 ⇥ 10�10

Azelaic acid C9H16O4 188.1049 [M-H]� (-4.2) 2.14 g L�1 1.151 ⇥ 10�10

cis-Pinonic acid C10H16O3 184.1099 [M-H]� (1.9) 7.8 g L�1 1.485 ⇥ 10�7

Myristic acid C14H28O2 228.2089 [M-H]� (0.7) 20.0 mg L�1 6.711 ⇥ 10�9

Palmitic acid C16H32O2 256.2402 [M-H]� (1.3) 7.2 mg L�1 7.175 ⇥ 10�10

Table F.1: Compounds used for method development. Water solubility data are fromYaws (2003), and vapor pressure is estimated by taking the average of predictionsfrom the ‘Evaporation’ (Compernolle et al., 2011) and ‘SIMPOL.1’ (Pankow andAsher, 2008) models.

F.2.2 Particle-into-Liquid Sampler (PILS)Detailed characterization of the Caltech PILS, which is based on a modification of

the original design of Weber et al. (2001), is described by Sorooshian et al. (2006).The chamber aerosol is sampled through a 1 µm cut size impactor with a flowrate of 12.5 L min�1, and passed successively through individual acid and base gasdenuders and an organic carbon denuder to remove inorganic and organic vapors.A steam flow generated at 100 �C is adiabatically mixed with the cooler sampledair in a condensation chamber, creating a high water supersaturation environmentin which particles grow su�ciently large (Dp > 1 µm) for collection by inertialimpaction onto a quartz plate. Impacted particles are transported to a debubblerby a washing flow (0.15 mL min�1) comprising 50% water and 50% isopropanol.The sampled liquid is delivered into vials held on a rotating carousel. This designis especially beneficial for the application of liquid chromatography for organiccompound separation and selection, which generally requires a longer time than thePILS collection cycle. Under the current configuration, a 5 min time resolution canbe achieved for the characterization of particle-phase dynamics.

F.2.3 UPLC/ESI-Q-TOFMSThe PILS collected samples are stored at 4 �C and analyzed by UPLC/ESI-Q-

TOFMS within 24 h without further pretreatment. A WATERS ACQUITY UPLCI-Class System, coupled with a Quadrupole Time-of-flight Mass Spectrometer (XevoG2-S QToF) and equipped with an electrospray ionization (ESI) source, was usedto identify and quantify the PILS collected samples, including organic aerosol

456

standards and chamber generated SOA. An ACUITY BEH C18 column (2.1 ⇥50 mm) was used to separate relatively nonpolar species, including vanillic acid,azelaic acid, cis-pinonic acid, myristic acid, palmitic acid, and IEPOX-derivedSOA. The eluent program is: 0-3.2 min: 100% A (0.1% formic acid in water);3.2-3.5 min: 10% A and 90% B (acetonitrile); 3.5-5 min: 100% A. The totalflow rate is 0.5 mL min�1 and the injection volume is 5 µL. For extremely polaror water soluble compounds, including adenine, adonitol, adipic acid, d-sorbitoland diethylmalonic acid, an ACUITY BEH Amide column (2.1 ⇥ 150 mm) wasused for species separation. The eluent program is: 0-3 min: 5% A (10 mMammonium formate in water at pH = 9) and 95% B (acetonitrile); 3-3.01 min:45% A and 55% B; 3.01-8 min: 5% A and 95% B. The total flow rate is 0.6 mLmin�1 and the injection volume is 4 µL. Note that the eluent program used here iscustomized for the characterization of test aerosols that are composed of a singlechemical standard. For chamber-derived SOA samples that might contain thousandsof species with a variety of polarities, key parameters of the eluent program, suchas the type and ratio of polar vs. nonpolar solvents, the additives and pH range inthe mobile phase, and the overall elution duration, need to be optimized to achievecompound specificity. The choice of separation modes is also very important.For extremely polar compounds, hydrophilic-interaction chromatography is a moresuitable approach, compared with reverse-phase chromatography, which has been,though, widely used in particle-phase speciation in previous studies. Optimumelectrospray conditions are: 2.0 kV capillary voltage, 40 V sampling cone, 30 Vsource o�set, 120 �C source temperature, 350 �C desolvation temperature, 30 L h�1

cone gas, and 650 L h�1 desolvation gas. MS/MS spectra were obtained by applyinga collision energy ramping program starting from 15 eV to 50 eV over one MS scanin the collision cell. Accurate masses were corrected by a lock spray of leucineencephalin (m/z 556.2771 [M+H]+). Data were acquired and processed using theMassLynx v4.1 software.

F.3 Method evaluationF.3.1 Simulation of Hygroscopic Growth of Organic Aerosols

The operational principle of the PILS is to grow particles in the presence ofsupersaturated water vapor to the size of micro-droplets that can be collected byinertial impaction. A key parameter that governs the extent of particle growth isthe water accommodation coe�cient (↵w), which is defined as the fraction of watermolecules that are taken up by the particles upon collision with the surface. Here we

457

investigate numerically the influence of ↵w on the predicted final size of the dropletsby condensational growth of particles.

The growth of a particle of diameter Dp in the presence of supersaturated watervapor is governed by (Seinfeld and Pandis, 2006):

dDp

dt=

4 · Dw · Mw

R · T · Dp · ⇢p

⇥p1w � ps

w(Dp,Ts)⇤· f (Kn, ↵w) (F.1)

where Dw is the water vapor di�usivity, Mw is the water molecular weight, R is thegas constant, T is the temperature, ⇢p is the particle density, Kn (=2�/Dp) is theKnudsen number, f (Kn, ↵w) is the correction factor for noncontinuum conditionsand imperfect accommodation, p1w is the water saturation vapor pressure far fromthe particle, and ps

w(Dp,Ts) is the water vapor pressure over the particle surface.Taking into account the vapor pressure variation due to the Kelvin e�ect, ps

w(Dp,Ts)becomes:

psw(Dp,Ts) = p0

w(Ts) · �w · �w · exp✓

4 · � · ⌫lR · T · Dp

◆(F.2)

where p0w(Ts) is the water saturation vapor pressure at the particle temperature Ts,

�w is the mole fraction of water in the particle phase, � is the water surface tension,⌫l is the water molar volume, and �w is the water activity coe�cient, which isassumed to be unity here. Note that particles in the PILS condensation chamber areassumed to behave as a dilute solution so that �w approaches its infinite dilution limit�w ! 1. This is reasonable considering that the water volume fraction exceeds 0.93in particles after 0.1 s of hygroscopic growth. The water activity coe�cient deviatesfrom unity (<1) in non-ideal organic-salt-water mixtures (Zuend et al., 2008, 2010),which is the case during the initial particle growth stage in the PILS condensationchamber. This leads to an overestimate of the water vapor partial pressure overthe particle surface, and consequently, slower simulated particle growth rate due towater condensation. In other words, the actual particle hydroscopic growth rate inthe PILS condensation chamber should be faster than the model simulations.

The particle surface temperature is determined by an energy balance on theparticle (Seinfeld and Pandis, 2006):

�T =�Hv · Dw · ⇢w

k 0a · T1

·Sw,1 � exp

✓�Hv · Mw

R · T1· �T

1 + �T+

4 · � · ⌫lR · T · Dp

◆�· f (Kn, ↵)

(F.3)where �T = (Ts � T1)/T1, �Hv is the latent heat of vaporization of water, ⇢w isthe density of water steam, k

0a is the e�ective thermal conductivity of air corrected

458

for noncontinuum e�ects, and Sw,1 = p1w /p0w(T1) is the environmental saturation

ratio. Equation F.3 needs to be solved numerically to determine the particle surfacetemperature that is required in the calculation of the particle growth rate.

Figure F.1 shows the predicted hygroscopic growth of particles, with initial di-ameters ranging from 20 nm to 800 nm, as a function of water accommodationcoe�cient (↵w). Particles with highly hydrophilic surface composition (↵w ! 1)can grow su�ciently large (Dp > 1 µm) to be collected by inertial impaction within1 s, the particle residence time in the PILS condensation chamber. Water vapormass transfer to less hydrophilic particles (↵w 10�2) is still su�ciently rapidto grow particles to droplets >1 µm size. Significant kinetic reductions in wateruptake result when the particle-phase constituents are strongly hydrophobic (↵w 10�3). The e�ect of initial relative humidity (RH) in aerosol sampling flow on theultimate droplet size becomes critical when ↵w is ⇠10�3. The simulations suggestthat PILS can potentially collect particles with a wide range of water solubility, dueto the substantial imbalance in water vapor abundance far from and over the particlesurface. As discussed earlier, the assumption of ideal water-ion-organic interactionsleads to the underestimation of particle hygroscopic growth rate. As a result, thespecification of ↵w in the order of ⇠10�3 as a criterion for "hydrophobic" particlesthat fail to grow over 1 µm diameter might be a bit conservative.

F.3.2 Collection E�ciencyThe PILS collection e�ciency may deviate from unity owing to three processes:

1) particle losses due to gravitational settling, di�usion, and inertial deposition dur-ing transport in the PILS plumbing system and condensation chamber, 2) evaporationof semi-volatile components during adiabatic mixing with steam, and 3) imperfectaccommodation of water vapor on particles, if hydrophobic (Orsini et al., 2003;Sorooshian et al., 2006; Weber et al., 2001). In the present study, we characterizethe PILS collection e�ciency by varying the particle chemical composition (i.e.,polarity and volatility) and size distribution. The overall mass collection e�ciency(CEPILS) can be obtained by comparing the mass concentration measured by theUPLC/ESI-Q-TOFMS and that derived from the DMA measured particle volumedistribution, assuming particle sphericity (Sorooshian et al., 2006):

CEPILS =1000 · m · Ql · DF · ⇢l

Qg · ⇢p · Vp(F.4)

where 1000 is the unit conversion factor, m is the UPLC/ESI-Q-TOFMS measuredorganic standard mixing ratio (ppm), Ql is the liquid sampling flow rate (1.5 mL

459

Figure F.1: Simulations of particle condensational growth over an initial size rangeof 20-800 nm diameter in the presence of supersaturated water vapor for di�erentassumed water accommodation coe�cients (↵w = 10�3, 10�2, and 10�1) at initialrelative humidities (RH) of 5% (upper panel) and 80% (lower panel) in the chamber.The black dashed line denotes the threshold value (1 µm in diameter) of particlesize for e�ective PILS collection.

min�1), DF is the dilution factor that accounts for the water vapor condensationon the impactor wall and air bubbles during the filling of vials, ⇢l is the densityof collected liquid, which is assumed as the density of the washing flow (0.893 gcm�3), Qg is the gas sampling flow rate (12.5 L min�1), Vp is the particle total volumeconcentration (µm3 cm-3) derived from the DMA measured number distribution,and ⇢p is the particle density (g cm�3). Note that ⇢p here is assumed as the densityof corresponding chemical standards that are used to generate test aerosols. Forchamber-derived SOA, ⇢p can be derived from the AMS measured O:C and H:Cratios (Kuwata et al., 2012). A 5 min o�set in the PILS measurement is taken intoaccount to retrieve the particle concentration at the moment of entry into the PILSinlet.

Figure F.2 shows the measured PILS mass collection e�ciency (CEPILS) asa function of particle water solubility and volatility. Consistent with the modelprediction, the high collection e�ciency in the PILS is achieved not only for highlywater soluble and non-volatile aerosols such as sorbitol, but also for slightly watersoluble and intermediate/semi-volatile aerosols such as pinonic acid. The watersolubility of particle-phase constituents clearly governs CEPILS. The e�ect of

460

Figure F.2: Measured PILS overall mass collection e�ciency for the test organicaerosols as a function of volatility, which is defined as the vapor pressure of com-pound i as pure liquid at 20 �C on the logarithm scale (log10P0

L,i, atm), and watersolubility, which is defined here as the maximum amount of the compound i thatwill dissolve in pure water at 20 �C on the logarithm scale (log10Hi, g L�1). Com-pound volatility is categorized according to intermediate volatility organic com-pound (IVOC), semi-volatile organic compound (SVOC), and low volatility organiccompound (LVOC). Water solubility is categorized broadly as insoluble (I), slightlysoluble (SS), and soluble (S).

volatility on CEPILS, however, is not discernable, indicating that the evaporation ofsemi-volatiles from particles during adiabatic mixing with steam is negligible forcompounds with vapor pressure <10�6 atm. We define the particle water solubilitythreshold of 1 g L�1 above which >0.6 mass collection e�ciency can be achieved.Considering that SOA water solubility data are generally unavailable, this quantitycan be related to the average particle O:C ratio, which is relatively well constrainedby the AMS measurement.

Figure F.3 shows the water solubilities (the maximum amount of the compoundi that will dissolve in pure water at 20 �C on the logarithm scale, log10Hi/g L�1)of a variety of organic compounds, including carboxylic acids, alcohols, carbonyls,esters, and ethers, as a function of their O:C ratios. Regardless of the nature offunctionalities in the molecule, the water solubility increases with the O:C ratio andeventually reaches a plateau, a miscible state with water. The O:C ratio correspond-

461

Figure F.3: Water solubility of carboxylic acid, alcohol, carbonyl, ester, and etherstandards at 20 �C as a function of their O:C ratios. Each data point representsan individual compound with certain functionalities. The horizontal gray linerepresents the water solubility threshold (1 g L�1) at which the PILS collectione�ciency of ⇠0.6 will be achieved. The two vertical gray lines define a region ofaverage particle O:C ratios that correspond to the water solubility threshold. Datasource: Yaws (2003).

ing to the water solubility threshold of 1 g L�1 varies with functionalities, rangingfrom 0.07 to 0.26. Considering that SOA is generally a mixture of species thatmight contain all the functional groups above, we suggest that the upper limit of theO:C ratio, 0.26, represents a criterion to constrain an e�ective PILS mass collectione�ciency. The AMS measured O:C ratios for SOA produced from isoprene + OH,↵-pinene + O3/OH, aromatics + OH, and C12-alkane + OH in the Caltech Envi-ronmental Chamber have been summarized previously (Chhabra et al., 2010, 2011;Loza et al., 2014; Zhang and Seinfeld, 2013). Among all the systems investigated,long-chain alkane derived SOA exhibits the lowest O:C ratio, ranging from ⇠0.2 to⇠0.3, over the course of 3-36 h experiments ([OH] exposure = 2-8 molecules cm�3

h). The average O:C ratios measured from other SOA systems, on the other hand,exceed the threshold, 0.26, at which >0.6 mass collection e�ciency by PILS canbe achieved. Thus, PILS is a candidate as a high e�ciency, time-resolved particlecollection method for a majority of SOA systems.

Particles are subject to gravitational settling, di�usion, and inertial deposition in

462

the PILS plumbing system and condensation chamber. The largest losses occur forsmall (Dp < 10 nm) and large particles (Dp > 1000 nm), as predicted to be 2.2% and8.6%, respectively (Sorooshian et al., 2006). Considering that the PILS transmissione�ciency (defined as the fraction of particles of a certain size that are transportedthrough the PILS plumbing system and condensation chamber) depends stronglyon the particle size, the impact of variations in the particle size distribution on thePILS overall mass collection e�ciency (CEPILS) is at issue. For a typical SOAchamber experiment, the initial size distribution of seed particles spans from ⇠20nm to ⇠600 nm, with a median diameter of ⇠60-70 nm. Growth driven by gas-phasephotochemistry and gas-particle partitioning occurs primarily on large particles and,as a result, the number median diameter shifts to ⇠200 nm after ⇠20 h of irradiation.To mimic the progression of the particle size distribution during SOA formation, testparticles were initially generated by atomizing concentrated pinonic acid solutioninto the chamber. An external flow pulse creates intense turbulent mixing in thechamber. As a result, the median diameter of the cis-pinonic acid particles shiftedfrom ⇠90 nm to ⇠200 nm during the 3.5 h experiment, due to rapid deposition ofthe smaller particles onto the chamber wall (see Figure F.4). The overall PILS masscollection e�ciency, on the other hand, remains consistently close to unity, within⇠8% uncertainty. This test demonstrates that change in particle size distributionduring SOA formation and evolution under typical chamber experimental conditionsdoes not significantly impact the PILS collection e�ciency.

F.3.3 Sensitivity, Detection Limit, and UncertaintiesThe extent of linearity of UPLC/ESI-Q-TOFMS detected signals for PILS col-

lected organic aerosols was examined in response to changes in particle mass load-ings. Sorbitol, azelaic acid, pinonic acid, and adipic acid aerosols, with massconcentrations ranging from 20 to 300 µg m�3, were generated by varying the at-omization and dilution duration. Mass concentrations of these four organic aerosolstandards were derived from the SMPS measured particle volume distribution. TheUPLC/ESI-Q-TOFMS signals for sorbitol, azelaic acid, and pinonic acid, as shownin Figure F.5, exhibit linearity (R2 = 0.99) with organic mass loadings. The sen-sitivity for adipic acid, however, is drifting low when particle concentrations arediluted, and as a result, there is a strong negative o�set in the linear fitting curve.The low sensitivity for adipic acid might be attributed to the slight solubility ofadipic acid dimer, which is the dominant particle-phase form of adipic acid (Wolfsand Desseyn, 1996).

463

Figure F.4: PILS overall mass collection e�ciency (upper panel) for cis-pinonicacid aerosols as a function of particle number distribution (lower panel). Over ⇠3h, aerosol wall deposition from intense turbulent mixing in the chamber caused thenumber median aerosol diameter to shift from 90 nm to 200 nm.

The detection limit of the PILS + UPLC/ESI-Q-TOFMS system can be estimatedfrom the expected UPLC/ESI-Q-TOFMS sensitivity and measured flow rates ofsampling air and liquid washing stream. The UPLC/ESI-Q-TOFMS detection limit(LOD) was calculated according to the expression:

SN=

ki · LOD�

(F.5)

where S/N is the signal-to-noise ratio, ki is the response factor or the sensitivity ofUPLC/ESI-Q-TOFMS for species i, and � is the standard deviation of the response.Using a signal-to-noise ratio of 2, LOD of 0.68 ppb is obtained for cis-pinonicacid as an example. With a gas flow rate of 12.5 L min�1, and a washing flowrate of 0.15 mL min�1, the PILS + UPLC/ESI-Q-TOFMS system is estimated tohave a detection limit of 19.02 ng m�3 for cis-pinonic acid in the particle phase.By analogy with other compounds that are selected by UPLC, the sensitivity ofthe Q-TOFMS technology can be practically achieved as low as ⇠ ppb level. Thus

464

Figure F.5: UPLC/ESI-Q-TOFMS measured mass concentration of sorbitol, azelaicacid, pinonic acid, and adipic acid in PILS samples versus the DMA-measured totalparticle mass. Linear fits to the data are shown with the mean R2 = 0.99.

we expect the detection limit of PILS + UPLC/ESI-Q-TOFMS measurement of themass of particle components in the range of tens of ng m�3 to a few µg m�3.

Uncertainties in the PILS + UPLC/ESI-Q-TOFMS measurement arise mainlyfrom variation of the collected liquid volume due to the existence of air bubbles.Specifically, the washing flow carries the impacted droplets via a stainless steelmesh wick on the perimeter of the impactor to a ’T-shape’ debubbler. Air bubblesare vacuumed out of the system by a peristaltic pump. The liquids are deliveredinto the injection needle that is inserted into individual vials by two syringe pumps.Perfect debubbling cannot always be achieved and as a result, the collected liquidvolumes in some vials can drift low. The measurement uncertainties due to theexistence of air bubbles in the liquid samples are estimated to be <11% by weighinga batch of vials over one carousel running cycle (72 vials). Repeatability in thedetected yield of ions from the electrospray ionization of the analyte is anotherpotential origin of measurement uncertainties. This was estimated to be <3% by

465

repeating injection of 4 µL standard organic samples into the electrospray source.The e�ect of ESI-MS sensitivity drift on the measurement accuracy is accounted forby regularly performing the repeatability of organic standards (one standard everyten samples) during routine sample analysis.

While the PILS + UPLC/ESI-Q-TOFMS system is demonstrated to achieve time-resolved molecular speciation of secondary organic aerosols, we acknowledge thepotential artifacts and uncertainties arising from aerosol sampling and analysis pro-cedures. First, the hydrolysis of labile functionalities, mainly including organicnitrates, carbonyls, and epoxides, upon solvation is inevitable in any aerosol col-lection, pretreatment and analysis procedures using water as the solvent. Second,water steam (100 ± 2 �C) is mixed with aerosol sample flow (25 ± 5 �C) in thePILS condensation chamber, with temperature eventually stabilized as ⇠36 �C. Dur-ing mixing, organic aerosol constituents are subjected to evaporation and thermaldecomposition. We have demonstrated earlier that the evaporation of semivolatileorganic compounds is not significant during adiabatic mixing (Figure F.2). Steadydecomposition of thermally labile compounds in neutral solution, such as organicperoxides, requires a temperature higher than ⇠80 �C (Hart, 1949). Consideringthat the residence time of the PILS condensation chamber is only 1 s, it is expectedthat the original structure of thermally labile organic molecules should be mostlyintact in such a short timescale.

F.4 Method Application: IEPOX Uptake onto Acidified and Hydrated Am-monium Sulfate Particles

The heterogeneous chemistry of isoprene epoxydiols (IEPOX), a second genera-tion product formed from isoprene photochemistry in the absence of NO, contributessignificantly to the SOA formation (Lin et al., 2014, 2012; Nguyen et al., 2014a;Paulot et al., 2009b; Surratt et al., 2010). Proposed mechanisms leading to SOAproduction involve the ring opening of the epoxide group, followed by the additionof available nucleophiles in the condensed phase, e.g., tetrol production via the ad-dition of water and organosulfate production via the addition of sulfate (Eddingsaaset al., 2010; Nguyen et al., 2014a; Surratt et al., 2010). The PILS + UPLC/ESI-Q-TOFMS method was employed to characterize the SOA formation and evolution viathe reactive uptake of IEPOX onto hydrated and acidified ammonium sulfate parti-cles. The IEPOX-derived SOA is a well-established system to test the performanceof the PILS + UPLC/ESI-Q-TOFMS technique.

466

F.4.1 Chamber experiment and instrument operation protocolsThe experiment was conducted in the Caltech 24 m3 Environmental Chamber

maintained at 20 �C. Detailed experimental protocols are presented in an earlierstudy (Nguyen et al., 2014a). The general experimental procedure is as follows:1) flushing the chamber with purified dry air for 24 h prior to the experiment; 2)humidifying the chamber to ⇠78% RH by passing purified air through a Nafionmembrane humidifier (FC200, Permapure LLC) that is kept wet by recirculationof 27 �C ultra-pure water (18 M⌦, Millipore Milli-Q); 3) Injecting inorganic seedaerosols by atomizing an acidified ammonium sulfate aqueous solution (0.06 M(NH4)2SO4 + 0.06 M H2SO4) followed by a custom-built wet-wall denuder; and 4)injecting the gas-phase trans �-IEPOX isomer by evaporating several droplets in aglass bulb with 6 L min�1 of purified and heated dry air (60 �C) for ⇠ 100 min.

The gas-phase IEPOX mixing ratio was monitored using a custom-modified Var-ian 1200 triple-quadrupole chemical ionization mass spectrometer (CIMS). In neg-ative mode operation, CF3O� was used as the reagent ion to cluster with the analyte[R], producing [R·CF3O]� or m/z [M+85]�, where M is the molecular weight of theanalyte. More details on the CIMS operation and data analysis are given by Zhanget al. (2015). Real-time particle mass spectra were collected continuously by anAerodyne High Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-AMS).The AMS switched once every minute between the high resolution ’W-mode’ andthe lower resolution, higher sensitivity ’V-mode’. The V-mode was utilized forquantification, as the higher m/z values exhibit a more favorable signal-to-noiseratio. The W-mode was used for ion identification and clarification. More detailson AMS operation and data analysis are given by Nguyen et al. (2014a).

F.4.2 Results and DiscussionsShown in Figure F.6 (upper panel) are the CIMS signals at m/z (-) 203, which

represents the fluoride cluster product of IEPOX (C5H10O3·CF3O�), and organicparticulate mass concentrations measured by the HR-AMS. Dominant ions observedin the HR-AMS spectra include m/z 29, m/z 43, m/z 53 (mostly C4H+5 ) and m/z 82(mostly C5H6O+). The latter two are considered as the tracers for IEPOX derivedorganic aerosols (Budisulistiorini et al., 2013; Lin et al., 2012). The time-dependenttrend of gas-phase IEPOX measured by CIMS during the injection period (0-100min) is almost identical to the AMS measured total particulate organics, indicatingthe uptake of IEPOX and the subsequent reactions are essentially instantaneous.Gas-particle equilibrium is rapidly established after the IEPOX injection. Over the

467

Figure F.6: Temporal profiles of normalized CIMS signals at m/z (-) 203, whichrepresents the fluoride cluster product of IEPOX (C5H10O3·CF3O�), and AMSmeasured total organic (upper panel), as well as IEPOX-derived sulfate ester anddimer measured by the PILS + UPLC/ESI-Q-TOFMS technique (lower panel).

remaining 100 min of the experiment, AMS measured total organic mass is at asteady state, whereas the CIMS signal at m/z (-) 203 exhibits a slow decay, which islikely a result of vapor wall losses of epoxide at high RH in the chamber.

For the PILS collected samples, the dominant ion observed in the UPLC/ESI-Q-TOFMS mass spectra in the negative mode is C5H11SO�

7 , which correspondsto the IEPOX-derived hydroxyl sulfate ester. The hydroxyl sulfate ester dimerC10H22SO�

10 is also observed in the form of [M-H]�. Corresponding normalizedmass spectra and temporal profiles of these two products are shown in Figure F.7and the lower panel of Figure F.6, respectively. Prompt formation of sulfate esterwas observed at the onset of IEPOX injection, and the time-dependent trend agreeswith the AMS measured organic aerosol growth curve. The sulfate ester dimer, onthe other hand, exhibits a 20 min induction period and creeps up over the course ofthe experiment. As authentic standards for IEPOX-derived organosulfates are notcommercially available, the UPLC/ESI-Q-TOFMS was externally calibrated with asurrogate quantification standard, d-galactone 6-sulfate sodium salt (Sigma Aldrich

468

Figure F.7: (Panel 1) Mass spectrum for the IEPOX-derived hydroxyl sulfate es-ter, detected as m/z = 215 in the form of C5H11SO�

7 ([M-H]�). (Panel 2) TheMS/MS fragmentation confirms the organosulfate with the m/z 97 (HSO�

4 ) daughterion. (Panel 3) Mass spectrum for the IEPOX-derived hydroxyl sulfate ester dimer,detected as m/z = 333 in the form of C10H20SO�

6 ([M-H]�).

> 98%), the elemental composition and functionality of which most closely resemblethat of the IEPOX-derived organosulfates. An ESI/MS response factor of 1.56 ⇥10�5 ppm/area, is obtained for the d-galactone 6-sulfate anion (C6H11SO�

9 ) at m/z259 and applied to quantify the IEPOX-derived organosulfate ester and dimer. Thissensitivity gives the overall equilibrium particle-phase organosulfate concentrationof ⇠12 µg m�3, which accounts for 17% of the AMS measured total organic masses.Note that although both belong to the organosulfate family and readily fragmentto bisulfate anion (HSO�

4 ) in the MS/MS analysis, the molecular properties of d-galactone 6-sulfate (C6H11SO�

9 ) and IEPOX-derived sulfate (C5H11SO�7 ) that a�ect

the ESI ionization e�ciency, such as pKa value, hydrophobicity, surface activity,etc., are not exactly the same. One should expect that the ESI-MS responses forthese two organosulfates di�er even under identical instrument operation conditions,and as a result, the conclusion that IEPOX-derived organosulfates account for 17%of the AMS measured total organic masses has a degree of uncertainty.

F.5 ConclusionsWe introduce here the combination of two well-established analytical techniques,

PILS and UPLC/ESI-Q-TOFMS, for time-resolved and quantitative measurement of

469

chamber-generated organic aerosol chemical composition. The PILS + UPLC/ESI-Q-TOFMS system shows its promising utility, including relatively high-time resolu-tion that allows for the investigation of aerosol dynamics and soft ionization to iden-tify the integral molecular structure of particle-phase components. Additionally, theincorporation of liquid chromatography allows the pre-separation of species prior tothe electrospray ionization, thus making the quantification of individual compoundsplausible.

The PILS collection e�ciency (CEPILS) towards a population of sub-micronparticles composed of pure chemical standards, including carboxylic acids, polyols,and amines, is estimated by simultaneously comparing the DMA vs. UPLC/ESI-Q-TOFMS measured total organic mass concentrations. The overall mass collectione�ciency exceeds 0.6 for particles with water solubility of >1 g L�1, which corre-sponds to an average O:C ratio of >0.26. The AMS measured O:C ratios for SOAproduced from photooxidation of a variety of VOCs (e.g., isoprene, toluene, m-xylene, ↵-pinene, and naphthalene) in chamber experiments exceed 0.3, for whichit is possible to characterize time-resolved chemical composition of these SOA sys-tems at the molecular level. Instrument sensitivity and linearity were tested usingsingle component organic aerosols generated from sorbitol, azelaic acid, pinonicacid, and adipic acid.

The PILS + UPLC/ESI-Q-TOFMS method is then applied to study the SOA for-mation driven by reactive IEPOX uptake onto hydrated and acidic ammonium sulfateparticles. For the first time, time-resolved traces of IEPOX-derived organosulfateester and dimer are observed. The temporal profile of the organosulfate ester isessentially identical to the AMS observed organic growth curve. The equilibriumorganosulfate concentration potentially accounts for a significant fraction of the over-all organic aerosol mass resulting from reactive IEPOX uptake. The combinationof PILS collection with UPLC/ESI-Q-TOFMS analysis o�ers a new approach fortime-resolved and quantitative characterization of aerosol constituents at molecularlevel.